AI Is Good For Counting Penguins
January 21, 2020
Artificial Intelligence has crept its way into almost every aspect of our lives. Now, its usefulness is being employed in Antarctica. The penguin populations there are at serious risk.
According to a 2019 study from the British Antarctic Survey, the largest Emperor penguin colony suffered unprecedented breeding issues. This has been an ongoing problem for the past three years, due in part to climate change.
Researchers estimate that by 2100, these penguins could disappear from the area for good. And in order to intervene and help alleviate their breeding problem, researchers need to first study the penguin populations.
The current method that is used to achieve this is to manually count them from camera traps. This can be tricky due to perspective distortion, penguins standing too close together or clustering, and diversity of camera angles.
But with the use of AI, this time consuming and inaccurate method of counting penguins, can be discarded completely. The tech is a new cloud counting solution from Gramener, an Intel AI Builder member and data science company.
Gramener uses an image image dataset of Antarctica’s penguin colonies from the Penguin Watch Project, which included images from over 40 locations. Working with Microsoft AI for Earth, the company’s researchers trained a deep learning model to count the penguins.
The model uses a density-based counting approach to approximate the number of penguins in clusters of different sizes from the images. This solution has been repurposed and benchmarked on Intel® Xeon® Scalable processors and the Intel® Optimization for PyTorch for optimized performance.
This type of AI tech could be a godsend for researchers. The labour intensive work could be left to the AI component, which will then allow researchers to devote more of their time to more valuable work.
Photo Credit:
Pixabay from Pexels
Intel Corporation
